Sovereign AI has become a rallying cry for governments worldwide, but the definition is shifting in unexpected ways. Rather than requiring nations to build all their AI technology domestically, experts and policymakers are now arguing that true sovereignty means retaining meaningful control over how systems operate, who can access them, and whether governments can modify or replace them without vendor permission.

What Is Sovereign AI, Really?

Algeria's newly approved roadmap offers one vision of sovereignty: developing domestic AI capabilities from the ground up. The North African nation has committed to training 30,000 AI engineers by 2030, establishing a dedicated scientific research center, and building high-performance computing infrastructure entirely within national borders. The strategy prioritizes four core pillars: using open-source AI models developed locally, maintaining domestic data storage systems, building national computing capacity, and strengthening protections for government AI applications.

But across the Atlantic, technology leaders are proposing a different framework. Red Hat, a major enterprise software company, argues that sovereignty should be defined by agency, choice, and control rather than geography alone. This distinction matters because it acknowledges a practical reality: most governments cannot realistically build every technology component domestically, nor should they necessarily try.

"Sovereignty is really about agency, choice and control, and less about geography. Acceptable dependency comes down to this: can a public body run, audit, and fix its own infrastructure without waiting on a supplier's permission?" said Jonny Williams, chief digital adviser for UK public sector at Red Hat.

Jonny Williams, Chief Digital Adviser for UK Public Sector, Red Hat

How Are Governments Building Sovereign AI Infrastructure?

Three distinct approaches are emerging across different regions, each reflecting different priorities and resources:

• Algeria's Domestic-First Model: Building comprehensive local capacity through workforce development, research institutions, and startup ecosystems. The country launched its first AI and cybersecurity startup cluster in April 2026 and graduated its first cohort of 105 AI specialists from its National Higher School of Artificial Intelligence in June.
• Portugal's Sovereign Cloud Strategy: Classifying government processes by sensitivity level and applying different sovereignty requirements accordingly. Processes are categorized as neutral, routine, critical, or strategic, with progressively stronger data control and security requirements.
• UK's Open-Source Control Model: Emphasizing the role of open-source software in retaining control over underlying dependencies. By standardizing on enterprise open-source platforms, public bodies can maintain code ownership even if vendors change their terms or restrict access.

Portugal's approach, outlined in its National Plan for Sovereign Cloud released in May 2026, reflects a pragmatic middle ground. Rather than requiring every government organization to use identical domestic systems, the plan allows flexibility based on actual risk levels while maintaining strategic control over the most sensitive data and operations.

Why Is This Debate Happening Now?

Recent geopolitical events have exposed vulnerabilities in relying on overseas technology providers. The UK's Science, Innovation and Technology Committee warned that the government risks losing access to critical AI models "at the whim of its partners" if it doesn't develop realistic sovereign capabilities. The committee specifically cited recent US restrictions on some AI models as evidence that allies cannot always be counted on to maintain access to frontier technologies.

This concern extends beyond AI. Portugal's National Plan for Sovereign Cloud explicitly mentions growing risks from cybercrime, extreme weather events, and electronic warfare as reasons to strengthen digital sovereignty and resilience. For governments, the stakes are high: digital infrastructure now underpins everything from healthcare to financial systems to national defense.

What Does Acceptable Dependency Actually Look Like?

The emerging consensus suggests that governments should evaluate their technology dependencies systematically rather than attempting complete self-sufficiency. Red Hat's framework proposes asking three key questions about any technology a government relies on: What do we depend on? Under what conditions? And for how long?.

Open-source software plays a central role in this approach. By using open-source platforms, public bodies gain transparency into the code they're running and retain the freedom to modify it, switch vendors, or maintain it independently if a supplier changes its business model. This applies across the entire technology stack, from AI applications down through operating systems and cloud platforms.

Algeria's emphasis on open-source AI models aligns with this philosophy, though the country is also investing heavily in proprietary domestic capabilities. Portugal's plan includes training at least 10% of public sector IT specialists in digital sovereignty by 2028 and at least 1,000 public managers and project leaders by 2030, recognizing that technical knowledge is itself a form of sovereignty.

What Comes Next for Governments?

The UK is moving forward with practical implementation. The Government Commercial Agency is launching G-Cloud 15, a consolidated procurement framework designed to simplify how government organizations purchase cloud technology and expand opportunities for smaller suppliers to compete. This represents a shift toward building a more diverse, resilient ecosystem rather than concentrating power with a few large vendors.

At the European level, the proposed EU Cloud and AI Development Act would establish four sovereignty assurance levels and a common procurement framework for public administrations across the bloc. The legislation aims to triple European data center capacity within five to seven years, creating infrastructure that can support sovereign AI development across the continent.

For Algeria, the shift from policy to implementation is underway. A joint monitoring committee will supervise the roadmap's execution according to a defined timeline, marking a transition from strategic planning to concrete programs and infrastructure deployment. The country's target of training 30,000 AI engineers by 2030 represents a significant expansion of its current specialist workforce and signals long-term commitment to building domestic expertise.

What these initiatives share is recognition that sovereignty in the AI era is not binary. It's not a choice between complete independence and total dependence. Instead, it's about building the knowledge, infrastructure, and legal frameworks that allow governments to make informed decisions about which technologies they can trust, which they must control, and which they can safely depend on from partners, as long as they retain meaningful agency over their own systems.